IP Library Granted Patent US 10,445,855
Granted Patent B2
US 10,445,855 · App. 15/302,846 · Granted Oct 15, 2019

Lung segmentation and bone suppression techniques for radiographic images

Inventors: David S. Harding (Austin, TX); Sridharan Kamalakanan (Dallas, TX); Satoshi Kasai (Hachioji, JP); Shinsuke Katsuhara (Koganei, JP); James H. Pike (Carrollton, TX); Muhammad F. Sabir (Allen, TX); Jeffrey C. Wehnes (Richardson, TX)
Assignees: iCAD, Inc.; Konica Minolta, Inc.
G06T3/0093A61B6/5211G06K9/3233G06K9/342G06K9/4647G06T7/0012G06T7/12G06T7/75G06K2209/051G06T5/002G06T2207/10116G06T2207/20036G06T2207/20076G06T2207/20116G06T2207/30008G06T2207/30061G06T2207/30064
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 10,445,855
App. No.
15/302,846
Granted
Oct 15, 2019
Kind
B2
Abstract

Lung segmentation and bone suppression techniques are helpful pre-processing steps prior to radiographic analyzes of the human thorax, as may occur during cancer screenings and other medical examinations. Autonomous lung segmentation may remove spurious boundary pixels from a radiographic image, as well as identify and refine lung boundaries. Thereafter, autonomous bone suppression may identify clavicle, posterior rib, and anterior rib bones using various image processing techniques, including warping and edge detection. The identified clavicle, posterior rib, and anterior rib bones may then be suppressed from the radiographic image to yield a segmented, bone suppressed radiographic image.

Claims (76)

1. A method for performing bone suppression, the method comprising:

receiving a radiographic image;

detecting bones in the radiographic image, the detected bones comprising one or a combination of a clavicle bone, a posterior rib bone, and an anterior rib bone;

suppressing the detected bones in the radiographic image to generate a bone suppressed image; and

filtering the radiographic image prior to detecting the bones in the radiographic images,

wherein filtering the radiographic image includes removing high contrast medical devices from the radiographic image, and removing a lung background trend from the radiographic image.

2. The method of claim 1 ,

wherein removing the lung background trend comprises:

generating a fitted lung surface by fitting a polynomial to a lung surface in the radiographic image; and

subtracting the fitted lung surface from the radiographic image.

3. The method of claim 1 , wherein detecting the bones in the radiographic image comprises:

warping the radiographic image to generate a warped radiographic image, wherein the clavicle bones are depicted with less curvature in the warped radiographic image than in the radiographic image; and

performing edge detection in the warped radiographic image to detect the clavicle bones.

4. The method of claim 1 , wherein detecting the bones in the radiographic image comprises:

generating an angle map from clavicle bones in the radiographic image;

warping the radiographic image in accordance with the angle map to generate a first warped radiographic image, wherein the clavicle bones are depicted with less curvature in the first warped radiographic image than in the radiographic image;

updating the angle map in accordance with the clavicle bones depicted in the first warped radiographic image;

re-warping the radiographic image in accordance with the updated angle map to generate a second warped radiographic image, wherein the clavicle bones are depicted with less curvature in the second warped radiographic image than in the first warped radiographic image; and

performing edge detection in the second warped radiographic image to detect the clavicle bones, and

wherein generating the angle map from the clavicle bones in the radiographic image comprises:

marking the clavicle bones in the radiographic image using a training data set;

generating angle measurements from the marked radiographic image; and

interpolating the angle map from the angle measurements.

5. The method of claim 1 , wherein detecting the bones in the radiographic image comprises:

generating angle maps for ribs in the radiographic image, wherein a different one of the angle maps is generated for each of the ribs;

warping portions of the radiographic image depicting the ribs in accordance with a corresponding one of the angle maps to generate a plurality of warped sub-images, wherein the ribs are depicted with less curvature in the plurality of warped sub-images than in the radiographic image; and

performing edge detection on the warped sub-images to detect the ribs.

6. The method of claim 1 , wherein suppressing the detected bones comprises:

selecting one of the detected bones to suppress from the radiographic image;

identifying a rectangular sub-image depicting the selected bone;

clipping, from the rectangular sub-image, edges of selected bone positioned outside a lung region;

processing the rectangular sub-image to reduce a curvature of the selected bone;

estimating a bone profile from the processed rectangular sub-image; and

suppressing the bone profile from the radiographic image.

7. A method for performing bone suppression, the method comprising:

receiving a radiographic image;

detecting bones in the radiographic image, the detected bones comprising one or a combination of a clavicle bone, a posterior rib bone, and an anterior rib bone; and

suppressing the detected bones in the radiographic image to generate a bone suppressed image,

wherein suppressing the detected bones comprises:

selecting one of the detected bones to suppress from the radiographic image;

identifying a rectangular sub-image depicting the selected bone;

clipping, from the rectangular sub-image, edges of selected bone positioned outside a lung region;

processing the rectangular sub-image to reduce a curvature of the selected bone;

estimating a bone profile from the processed rectangular sub-image; and

suppressing the bone profile from the radiographic image, and

wherein estimating the bone profile from the processed rectangular sub-image comprises:

removing a background signal from the processed rectangular sub-image, thereby generating a background-removed image;

identifying upper and lower edge profiles of the selected bone depicted by the background-removed;

estimating the bone profile in accordance with the upper and lower edge profiles; and

estimating a bone edge signal for the selected bone based on the processed rectangular sub-image or the background-removed signal, wherein estimation of the bone edge signal is performed separately from estimation of the bone profile,

wherein estimating the bone edge signal for the selected bone from the processed rectangular sub-image comprises:

obtaining a directional subsampled sub-image of the selected bone by subsampling the processed rectangular sub-image at a higher frequency in a direction that is parallel to the selected bone than in a direction that is perpendicular to the selected bone; and

estimating the bone edge signal from the directional subsampled sub-image,

wherein estimating the bone edge signal from the directional subsampled sub-image comprises:

morphologically processing the directional subsampled sub-image to obtain a processed directional subsampled sub-image;

smoothing the processed directional subsampled sub-image to obtain a smoothed directional subsampled sub-image; and

estimating the bone edge signal from the smoothed directional subsampled sub-image, and

wherein smoothing the processed directional subsampled sub-image comprises:

smoothing regions the processed directional subsampled sub-image using a smaller sigma value for regions within a threshold distance of a bone-edge than regions greater than the threshold distance from the bone-edge.

8. A method for performing bone suppression, the method comprising:

receiving, by a processor, a radiographic image;

performing, by the processor, lung segmentation on the radiographic image to generate a segmented lung image;

detecting, by the processor, clavicle bones in the segmented lung image;

detecting, by the processor, rib bones in the segmented lung image; and

suppressing, by the processor, the detected bones from the radiographic image to generate a bone suppressed image,

wherein detecting clavicle bones in the segmented lung image comprises warping the segmented lung image to generate a warped image, wherein the clavicle bones are depicted with less curvature in the warped image than in the segmented lung image; and performing edge detection in the warped image to detect the clavicle bones.

9. The method of claim 8 , wherein performing lung segmentation on the radiographic image to generate a segmented lung image comprises:

identifying, by the processor, region of interest (ROI) boundaries within the radiographic image;

identifying, by the processor, lung boundaries in accordance with the ROI boundaries; and

merging, by the processor, the lung boundaries to generate a segmented lung image.

10. The method of claim 8 , wherein detecting the rib bones in the segmented lung image comprises:

generating angle maps for the rib bones in the radiographic image, wherein a different one of the angle maps is generated for each of the rib bones;

warping portions of the radiographic image depicting the rib bones in accordance with a corresponding one of the angle maps to generate a plurality of warped sub-images, wherein the rib bones are depicted with less curvature in the plurality of warped sub-images than in the radiographic image; and

performing edge detection on the warped sub-images to detect the rib bones.

11. The method of claim 8 , wherein suppressing the detected bones from the radiographic image to generate the bone suppressed image comprises:

iteratively suppressing each of the detected bones on a bone-by-bone basis.

Assignments (4)
SECURITY INTEREST Recorded Mar 30, 2020
From: ICAD, INC.
To: WESTERN ALLIANCE BANK
Reel/Frame 052266/0959 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 11, 2016
From: HARDING, DAVID S.; KAMALAKANAN, SRIDHARAN; PIKE, JAMES H.; SABIR, MUHAMMAD F.; WEHNES, JEFFREY C.
To: VUCOMP, INC.
Reel/Frame 040315/0833 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 11, 2016
From: KASAI, SATOSHI; KATSUHARA, SHINSUKE
To: ICAD, INC.; KONICA MINOLTA, INC.
Reel/Frame 040315/0987 →
MERGER Recorded Oct 11, 2016
From: VUCOMP, INC.
To: ICAD, INC.
Reel/Frame 039988/0412 →
Continuity (2)
Provisional Application 61976971 · Apr 8, 2014
Related Publication 20170032535A1 · Feb 2, 2017